Durable attitude change using active processing: A longitudinal study on the effects of AI dialogue

Abstract

This study investigates whether AI-driven conversations can produce durable attitude change toward controversial policies, specifically legal kidney markets and free trade. In Study 1, participants interacted with a GPT-3.5 chatbot using utilitarian, deontological, emotional, and narrative arguments. Attitudes were measured immediately, one week, seven weeks, and eighteen weeks after the intervention. All strategies shifted attitudes in the short term, with most changes remaining durable after one week. However, only narrative arguments remained effective at seven weeks, with aggregate change persisting at eighteen weeks. Neither argument type nor decision-making style interacted to affect persuasiveness. Study 2 found in-group/out-group messaging effective but equivalent. Vivid memory of AI interactions predicted positive attitude change, while accurate memory of the argument had no effect. These findings suggest that conversational AI can produce long-term attitude change through active cognitive processing, demonstrating its potential as a scalable tool for shifting public attitudes on complex issues through interactive engagement.